How come DDR3 is better than DDR2 and DDR2 better than DDR

[dv8silencer]

I am quoting from this Tom's Hardware article:

Consider the latency ratings of the three most recent memory formats: Upper-midrange DDR-333 was rated at CAS 2; similar-market DDR2-667 was rated at CAS 4 and today's middle DDR3-1333 is often rated at CAS 8. Most people would be shocked to learn that these vastly different rated timings result in the same actual response time, which is specifically 12 nanoseconds.

Because cycle time is the inverse of clock speed (1/2 of DDR data rates), the DDR-333 reference clock cycled every six nanoseconds, DDR2-667 every three nanoseconds and DDR3-1333 every 1.5 nanoseconds. Latency is measured in clock cycles, and two 6ns cycles occur in the same time as four 3ns cycles or eight 1.5ns cycles. If you still have your doubts, do the math!

So comparing the mentioned DDR2-667 with CAS 4 to DDR3-1333 with CAS 8, both will produce memory after a request in the same amount of time. For example, in one second, the DDR2-667 memory will have gone through 667 million cycles. Based on the CAS of 4 cycles, that will produce 667/4= 166 million units of memory in one second. For the DDR3-1333 with CAS of 8, in one second the memory will have gone through 1.333 billion cycles. Based on the CAS of 8 cycles, that will produce 1,333/8 = 166 million units of memory. Does this make sense? I am using my very beginner-level knowledge to understand this.
My question: How come DDR3-1333 memory outperforms DDR2-667 memory if they both produce the same amount of memory in the same amount of time? Also, using the same reasoning, how come DDR2-667 memory outperform DDR-333 memory? The only good the higher frequencies do is to mask their higher CAS timings? I'm probably missing a big point here.

I would like some input to clear this puzzle up.

Thanks

 

[jitpublisher]

You just asked the 1 million dollar question, and I think you answered it yourself pretty well.
I don't think that DDR2 is any faster than standard DDR, and DDR3 is any faster than DDR2 for the exact reasons you just mentioned. Sure, the Mhz goes up but at the drawback of higher latencies. Maybe when they are able to drop the latencies......?
I think it is all about marketing and selling something new.
Kind of like the 8800GT's suddenly popping up with 1 gig of memory for a hefty price. There is no way an 8800gt could even begin to use that much memory, but I have witnessed first hand people grabbing up them up left and right at local retailers proclaiming "Wow! This one has a gig of memory, it MUST be the best card here!"

 

[xx12amanxx]

Well from what ive read ddr3 is not going to give you any noticable performance increase over say ddr2..Seem's that the general consensous is if you have ddr2-800 going up to ddr3 or ddr2-1066 isnt going to make a difference. But going from ddr2-533 or so to ddr2-800 will provide a performance boost you can notice.

 

[mdaubert]

You are correct that DDR2-667 with CAS 4 will produce the same amount of data as DDR3-1333 with CAS 8. Until DDR3 with tighter timings are produced, there will be no big performance difference between DDR2 and DDR3. The reason for the shift, however, is because FSB speeds are going up and there is a performance loss associated with having your root memory clock slower than your system clock.

As a simple example, take a Pentium 4 running at 2.4GHz (yes, a bit dated but easy numbers). That's 200MHz system clock with a 12.0x multiplier = 2.4GHz. The 200MHz system clock quad-pumped gives you 800MHz of FSB bandwidth. The minimum speed DDR2 memory that can fully utilize 800MHz of bandwidth is DDR2-800 (200Mhz memory clock times 2 (DDR) times 2 (dual channel) = 800M ops/sec). Now consider that FSB speeds are rising to push more performance out of processors. DDR2-1066 is required to fully utilize a 1066MHz FSB. Getting DDR2 to align with a 1333MHz FSB is pushing the limits, DDR3 to the rescue. DDR3-800 = 100MHz memory clock * 4 (QDR) * 2 (dual channel). DDR3-800 can provide 800M ops/sec while operating at a 100MHz memory clock - half the clock required to drive the DDR2. Push the memory clock to 200MHz (the same as DDR2-800) and you get twice the bandwidth... DDR3-1600. The QX9775 runs with a FSB of 1600MHz... you'll have a hard time finding DDR2 that can keep up!!!

Hope this makes a little bit of sense.

 

[dv8silencer]

So what you're saying is, the DDR3 is better in the case of higher FSB just because it matches the frequency better, even though it still will produce data the same rate as a DDR2 will?

 

[lasting_omen]

Yes, that’s the jist of it. DDR2 can only go so fast to match an over clocked FSB. In order to get to the maximum over clock potential from an enthusiast motherboard and\or processor, faster memory is needed, hence DDR3. What is interesting is that I’ve found very few benchmarks that show faster memory with looser timings performing better. 90-95% of the test show slower memory speeds with tighter latencies wins the race. So in summary, you are paying more money for equal or slight lower memory performance with DDR3, unless you over clock a system that allows the FSB to go beyond the reach of enthusiast DDR2 memory. I'm guessing this will change when the memory controller is moved onboard future processors, I'll have to shake the 8-ball to get an answer on that one!

 

[e36_Jeff]

if I am not mistaken, i do believe the timings dictate things like how long the memory takes to respond to a given request, but do not dictate the transfer rates. the increased frequency of DDR2 and DDR3 does allow for higher transfer rates once the request has been make and carried out. also, DDR3 operates at lower voltages than DDR2 and DDR2 is lower than DDR, which lowers overall power draw.

 

[reconx86]

The REAL reason why they produced DDR3 is because the manufacturing process is cheaper, thats the REAL truth. The resy is just BS in some cases DDR2 outperforms DDR3. The RAM manufacturers wont tell you that, they only think about profit